AI in Marketing Has Finally Started Making Its Own Calls

A while back I came across a blog post that talked up today's marketing tools like something out of a sci-fi movie.

AI running customer journeys on autopilot. Agents writing your copy and picking your channels for you. Ads optimizing themselves around the clock. Read it and you'd think the marketing department could all take vacation at once and just hand the keys to the company.

But the moment you set foot inside a real marketing department, the story falls apart.

The data is dirty. The systems don't talk to each other. Ops looks at what AI suggests and has no idea what to do with it, so they fall back to Excel and write the plan by hand. The so-called "fully automated" thing is mostly just a slide-deck buzzword.

But here's what's interesting: this time, beneath the froth, there's something real.

What Does It Mean for AI to "Make Its Own Calls" in Marketing?

I recently went out and got a ground-truth look at five of the leading platforms in this space, side by side. Not by watching their launch keynotes — by asking them: of the customers who are actually paying you, how is it really going?

So what does "make its own calls" mean?

Put simply: AI used to be an advisor in marketing. You'd hand it a batch of users; it would score them — this one's likely to churn, that one's likely to buy. After scoring, the rest was still up to the marketers. Which channel, what time, how to rewrite the copy, how to set up the A/B test — all human.

Now, AI is starting to act as the executor.

It doesn't just tell you "this user is at risk of churning." It decides on its own: for this user, I'll send a push, not an email; 9 p.m., not 3 p.m.; warm copy, not a promo push; after sending, watch the reaction, and if they don't open, swap in a new version 12 hours later — the whole loop runs itself.

It can already do six things: pick the audience, pick the channel, pick the timing, drive the journey, adjust the creative, and run A/B tests. Three years ago, each of those was the job of one analyst plus one ops person.

This is the key shift in this wave of AI marketing: it's gone from scoring users to actually doing the work.

So the Real Question: Is Anyone Actually Using It?

This is what I wanted to know most. You can hype it all you want — if nobody's using it, it's just empty noise.

The numbers I got were pretty interesting. Of these five platforms, two told me that roughly 26% to 50% of their customers are actually using the AI decisioning features — early-to-mid stage. The other three went bigger, reporting 51% to 75% directly.

What does that mean?

Put the two sets together and the industry has already cleared the "early adopter" hump, and is sitting right at the inflection where the early majority is pouring in. AI in marketing isn't a toy for a few cutting-edge companies anymore — it's inside the daily workflow of a real slice of the market.

But about half of the customers still aren't using it.

That's an interesting window: the ones using it are already making money; the ones who aren't are still standing still.

For the Ones Using It — What Are They Actually Getting Out of It?

The specific numbers differ across vendors, but the direction is strikingly consistent. I pulled together the benefits they kept mentioning; there are roughly six:

• Faster launches. A campaign that used to take weeks to go from plan to live now takes days, sometimes hours.
• Higher conversion. Better audience selection, better timing, better channels — conversion naturally climbs.
• Stickier retention. The system smells churn coming early and auto-triggers a win-back; the churn curve flattens out.
• Smarter spend. AI doesn't spend by gut the way humans do; it knows which dollar returns more and piles budget there.
• Sharper targeting. Used to be gut calls; now it's data down to the individual.
• Faster payback. Time is money; stack the previous five together and the payback period visibly shrinks.

But what made me feel "this is really happening" wasn't any single-point win. It was one word — compounding.

What does that mean?

Once decisioning is automated, the system starts teaching itself. Every push you send today becomes training data for tomorrow. The more it runs, the sharper it gets; the sharper it gets, the more it runs. That compounding effect will eventually show up on the revenue curve — but only on one condition: you can't stop at "saving time." You have to actually let AI make decisions, not just have it run errands for you.

If It's So Great, Why Isn't Everyone Using It?

This is the part of the whole story most worth thinking through.

After talking to these platforms, I got a strikingly uniform answer: what kills AI decisioning is never the AI itself — it's the foundation the AI can't run on.

Where does the foundation give way? Four places.

First, data.

This is the single biggest hurdle every vendor mentioned, in unison. AI is a data-eating monster; feed it data that's dirty, fragmented, or stale, and the decisions it spits out will be crooked. Garbage in, garbage out. One vendor said it to me directly: of their customers who use AI well, every single one — no exception — first built a solid data foundation. No good data, no AI.

Second, skills.

The tech may be in place; the people might not be. AI gives you a nudge — "suggest switching channels" — and your ops team has to know: is this recommendation any good? Should we take it? If we take it, how do we land it? A lot of teams freeze. It's not that AI is dumb; it's that nobody in the room can actually harness it.

Third, goals.

What do you actually want the AI to do? Far fewer people have thought this through than you'd imagine. One platform told me they've seen too many customers buy the system with zero internal alignment — are we going for acquisition, for repeat purchase, or for cost reduction? With unclear goals, the AI just spins its wheels. The capability is there, but no direction is given, so the car still won't move.

Fourth, trust.

This one is the most hidden, and the most lethal.

AI says "this user is going to churn next week, push a coupon right now." Do you believe it?

A lot of marketers don't. And why not? Because the AI can't give them a reason. It just slaps a conclusion on the table and walks off. That's like a new hire banging on your desk every day saying "Boss, we have to do it this way" — and never explaining why. Sooner or later, you'll show them the door.

So a consensus is starting to surface in the industry: a good AI decisioning system can't just hand you an answer; it has to show you the reasoning. It has to let you understand it, challenge it, correct it. AI isn't thinking for you; it's thinking with you.

Data, skills, goals, trust. These four things have basically nothing to do with AI — and yet they decide whether AI can actually run inside your company.

What's really gumming up the AI is never the AI itself — it's everything around it.

So Where's the Money Going in 2026?

When the conversation turned to investment direction, the five platforms were singing basically the same tune. I grouped it into three buckets.

First bucket: real-time data.

The faster AI runs, the fresher the data has to be. Have it make real-time decisions while the user-behavior data is an hour old, and the decision is a joke. So everyone is building faster customer-signal pipelines, so that what AI gets is always the context of right now.

Second bucket: autonomous decisioning engines.

Predictive models are getting an upgrade. It used to be "predict who's going to churn"; now it's "after predicting, automatically launch the win-back flow — no human in the middle." The AI system is starting to behave like a self-adjusting flywheel, tweaking itself on the fly.

Third bucket: building trust between humans and AI.

This one honestly surprised me. It turns out people are seriously investing in training, in ROI frameworks, in "making AI less of a black box." Because they've figured something out: AI isn't a one-time deployment. It has to keep being trusted to keep being used.

Stack those three together and it really comes down to one sentence: in 2026, more than half of investment isn't going into AI itself — it's going into the ring of things around the AI.

The Real Inflection: From "Giving Advice" to "Doing the Work"

Put all the threads above together and you can see a clear inflection point —

2026 is the dividing line where AI in marketing goes from "advisor" to "executor."

Four shifts are happening at once.

One: from "advice" to "action."

AI used to say "I suggest you do this." The AI of the future says "I've already done it — take a look and see if you want to adjust." An extreme version: from content generation, to distribution path, to performance optimization, AI closes the loop end-to-end on its own; the human is just watching from the side.

Two: from "point-in-time tests" to "continuous experimentation."

A/B testing used to be a discrete event — plan it, run it, read the result, close it out. In the future, there is no standalone "A/B test" anymore; it's embedded in the system, running constantly, auto-selecting the winner, and pushing the winner front and center.

Three: from "periodic optimization" to "real-time recalibration."

It used to be: review once a week and tweak the parameters. Now: every click, every swipe a user makes, the system is micro-adjusting its next move. Every interaction becomes fuel for the next decision.

Four: from "external automation" to "in-product intelligence."

AI decisioning is starting to move inside the product itself. How a user is guided inside the app, what features they see, when to push what content — all decided by AI based on their real-time behavior. AI is slowly blurring the line between marketing and product.

Stack those four shifts together and it means one thing —

The marketer's role is shifting from "the person who builds campaigns" to "the person who manages the AI that builds campaigns."

It's no longer you personally configuring every workflow, writing every version of the copy, running every test. It's you setting the goals, drawing the boundaries, watching the AI do the work, and stepping in to correct course when you need to.

There's a clean analogy for this: the control tower.

The controller in an airport tower doesn't fly planes themselves. But when each plane takes off, when each one lands, which route each one flies — all of it happens under their direction. They're not pilots, but without the tower, the whole airport descends into chaos.

The marketer of the future is the AI's air-traffic controller.

One Honest Closing Thought

After this whole round of conversations, my biggest takeaway is this —

Don't get fooled by those "AI can already self-drive marketing" headlines on the blogs, and don't get numbed by the old "AI is still early" line inside your company.

The truth is in the middle.

AI in marketing can already do a lot — but whether it does it well depends on your data, your people, the clarity of your goals, and the room for trust you give it. This is an engineering problem, not a magic problem.

So if you run a marketing team, rather than agonizing over "should we get on board with AI," you're better off doing three things first —

Build a solid data foundation. Unify sources, clean the structure, keep the signals flowing. Without this step, nothing else is even worth discussing.

Get your team comfortable working with AI. It's not enough to know how to use the tools — your team has to be willing to challenge its output and redirect its course.

Rethink the division of labor in marketing. Which decisions should be handed off, which must stay in human hands, what counts as "enough," and what counts as "too far."

The essence of this whole thing was never about whether AI replaces marketers.

It's about whether you can go from being the person flying the plane with your own hands to the person standing in the tower directing one plane after another.

Some people started making that turn a year ago. Some are still standing in place.

To be honest, I haven't figured out either where this plane is going to end up flying. But one thing is certain — the seats in the tower are getting fewer.